In the present study, in order to evaluate the feasibility of personalized chemotherapy, a prospective randomized pilot study was performed in 30 advanced or recurrent gastric cancer patients. As we have demonstrated previously, the expressions of mRNA from tumor biopsy samples for seven molecular markers, i.e., dihydropyrimidine dehydrogenase (DPD), glutathione S-transferase (GST)-pi, beta-tubulin (tub), O6-methylguanine-DNA methyltransferase (MGMT), multiple drug-resistant protein (MRP)-1, NADPH/quinone oxidoreductase (NQO)-1, and cytochrome p450 (P450), were examined by reverse transcription-polymerase chain reaction (RT-PCR) analysis and therapy was recommended in a flow chart that depended on the level of expression of these predictive molecular markers. We chose 12 therapeutic plans, including best supportive care (BSC). We treated 15 patients according to the gene expression profiles, and the remaining 15 patients (controls) were treated without recommended regimens, and the therapy was continued after the expression profiles were checked. Interestingly, 11 of 26 lesions (42.3%) responded after treatment given according to gene expression analysis; however, no clinical response was detected in the control group. The prediction of the response, including resistance, was successful in 75.9% by the gene expression profiles. Moreover, the survival of the patients with the recommended treatment was better than that of patients without a recommended protocol. These results indicate that personalized treatment may be beneficial for gastric cancer chemotherapy and further randomized trials should be carried out in the future.